Energy Fuels XXXX, XXX, 000–000 : DOI:10.

1021/ef901240p

Forecasting World Crude Oil Production Using Multicyclic Hubbert Model

Ibrahim Sami Nashawi,*,† Adel Malallah,† and Mohammed Al-Bisharah‡
†
Department of Petroleum Engineering, College of Engineering and Petroleum, Kuwait University, P.O. BOX 5969, Safat 13060,
Kuwait, and ‡Kuwait Oil Company, P.O. Box 9758, Ahmadi 61008, Kuwait

Received October 29, 2009. Revised Manuscript Received December 31, 2009

The year 2008 has witnessed unprecedented fluctuations in the oil prices. During the first three-quarters, the oil price
abruptly increased to $140/bbl, a level that has never been reached before; then because of the global economic
crisis, the price dramatically plunged to less than $50/bbl by the end of the year losing more than 64% of the
maximum price in less than three months period. The supply of crude oil to the international market oscillated to
follow suite according to the law of supply and demand. This behavior affected oil production in all exporting
countries. Nonetheless, the demand for crude oil in some developing countries, such as China and India, has
increased in the past few years because of the rapid growth in the transportation sector in addition to the absence of
viable economic alternatives for fossil fuel. The rapid growth in fuel demand has forced the policy makers world-
wide to include uninterrupted crude oil supply as a vital priority in their economic and strategic planning.
Even though forecasting should be handled with extreme caution, it is always desirable to look ahead as far as
possible to make an intellectual judgment on the future supplies of crude oil. Over the years, accurate pre-
diction of oil production was confronted by fluctuating ecological, economical, and political factors, which
imposed many restrictions on its exploration, transportation, and supply and demand. The objective of this
study is to develop a forecasting model to predict world crude oil supply with better accuracy than the existing
models. Even though our approach originates from Hubbert model, it overcomes the limitations and restric-
tions associated with the original Hubbert model. As opposed to Hubbert single-cycle model, our model has
more than one cycle depending on the historical oil production trend and known oil reserves. The presented
method is a viable tool to predict the peak oil production rate and time. The model is simple, accurate, and
totally data driven, which allows a continuous updating once new data are available. The analysis of 47 major
oil producing countries estimates the world’s ultimate crude oil reserve by 2140 BSTB and the remaining
recoverable oil by 1161 BSTB. The world production is estimated to peak in 2014 at a rate of 79 MMSTB/D.
OPEC has remaining reserve of 909 BSTB, which is about 78% of the world reserves. OPEC production is
expected to peak in 2026 at a rate of 53 MMSTB/D. On the basis of 2005 world crude oil production and
current recovery techniques, the world oil reserves are being depleted at an annual rate of 2.1%.

1. Introduction of production history. Decline curve analysis,1,2 black oil

The demand for crude oil started to accelerate with the
invention of the internal combustion engine in the late nine-
teenth century to become one of the most important commodi-
ties traded worldwide. Modern civilization, as it is known to us,
heavily depends on crude oil and its byproducts. So far, the
industrialized nations have completely taken for granted an
uninterruptable supply of cheap hydrocarbon.
One of the prime objectives of the petroleum industry is to
provide the modern world with continuous flow of hydro-
carbon fluids, oil and gas, while making a profit. Petroleum
liquids are exhaustible resources; thus, a good forecasting
scheme of oil supply will be crucial to all parties involved in the
petroleum business, such as oil companies, financial institu-
tions, public policy planners and makers, and oil exporting
and importing countries. Such a model will also help bring
stability and security to the crude oil market.
In the past, many researchers have developed different
methods to forecast future production of crude oil using either
available determinations of ultimate reserves or extrapolation
*To whom correspondence should be addressed. Telephone: þ3 965
24987570. Fax: þ3 965 24836058. E-mail: is.nashawi@ku.edu.kw.
(1) Mannon, R. W. Presnted at the SPE Annual Fall Meeting,
Denver, CO, October 1965; Paper SPE 1254.
model history matching, and past trend extrapolations are
often considered statistical methods of production fore-
casting.3-8 Econometric models based on physics, economics,
technology, and remaining reserves were also used.9-11 In
1995, Skov12 presented a discussion of the various prediction
tools used to forecast energy supply and demand. More
recently, Monte Carlo simulation models13 and artificial
(2) Arps, J. J.; Mortada, M.; Smith, A. E. J. Pet. Technol. 1971, 23 (6),
671-675.
(3) Campbell, C. J. The Coming Oil Crisis; Multi-Science Publishing
Co. and Petroconsultants S.A.: Brentwood, U.K., 1997; p 86.
(4) Hubbert, M. K. Drilling Prod. Pract. 1956, 7–25.
(5) Hubbert, M. K. Energy Resources; Report to the Committee on
Natural Resources, Publication 1000-D, Natl. Academy of Science/Natl.
Research Council: Washington, D.C., 1962.
(6) Hubbert, M. K. AAPG Bull. 1967, 51, 2207–2227.
(7) Hubbert, M. K. Techniques of Prediction as Applied to Production
of Oil and Gas; Proceedings; NBS Special Publication 631; U.S. Dept. of
Commerce: Washington, DC, 1982.
(8) Laherrere, J. H.; Sornette, D. Eur. Phys. J. 1998, 525–539.
(9) Taylor, P. J. J. Pet. Technol. 1997, 49 (5), 502–507.
(10) Cleveland, C. J.; Kaufmann, R. K. Energy J. 1991, 12 (2), 1–30.
(11) Kaufmann, R. K. Resourc. Energy 1991, 13, 111–127.
(12) Skov, A. M. Presented at the SPE Hydrocarbon Economics and
Evaluation Symposium, Dallas, TX, March 1995; Paper SPE 30058.
(13) Garcia, A.; Mohaghegh, S. D. Presented at the SPE Eastern
Regional Conference and Exhibition, Charleston, WV, September 2004;
Paper SPE 91413.

r XXXX American Chemical Society A pubs.acs.org/EF

Energy Fuels XXXX, XXX, 000–000 : DOI:10.1021/ef901240p
14,15
intelligence tools, such as fuzzy logic and neural networks,
were also used for this purpose.
The Hubbert4-7 model is one of the most renowned
statistical models for the prediction of oil and gas produc-
tion. Initially presented in 1956, Hubbert4 fitted bell-shape
curves to cumulative production and discoveries to forecast
oil production in the United States (U.S.). He predicted that
oil production in the U.S. lower 48 states would ultimately
produce about 170 BSTB and that production would peak
in 1970. At that time, these predictions were very pessimistic
compared to those presented by famous forecasting agen-
cies such as U.S. Geological Survey (USGS) and other
forecasters.16-18 However, time had shown that Hubbert
forecasts were remarkably accurate; oil production did
indeed peak in 1970. Since that time, the Hubbert model
gained worldwide popularity because of its simplicity and
availability of required data and was extensively tested and
used to forecast oil production worldwide.19-27 Having
compared the forecast results of various methods to those
of Hubbert, Cleveland and Kaufmann10 praised the Hub-
bert model, stating that, despite its lack of theoretical basis,
its symmetric parabola predicts production more accu-
rately than regression curves, economic models, or delphi
techniques.
Initially, Hubbert4 obtained his prediction empirically.
He fitted historical production data by a normal or Gauss
bell-shape curve by making two major assumptions: (i)
initially the production rate must start at zero, increase to
a maximum, then decline to zero and (ii) the area under the
production curve is equal to the ultimate oil recovery as time
approaches infinity. Later, Hubbert5-7 presented the math-
ematical foundation for his model where a logistic curve was
used to fit cumulative oil production versus time. In the past,
several authors19-21,28-30 have illustrated that Hubbert
model with only one full cycle is appropriate to forecast
crude oil production for some countries, where the produc-
(14) Weiss, W. W.; Balch, R. S.; Stubbs, B. A. Presented at the SPE/
DOE Improved Oil Recovery Symposium, Tulsa, OK, April 2002; Paper
SPE 75143.
(15) Al-Fattah, S. M.; Startzman, R. A. SPE Prod. Facil.s 2003, 18 (2),
84–91.
(16) Zapp, A. D. U.S. Geol. Surv. Bull. 1962No. 1142-H.
(17) Duncan, D. C.; McKelvey, V. E. Table-9 of United States
Resources of Fossil Fuels, Radioactive Minerals, and Geothermal Energy;
Research and Development on Natural Resources, Federal Science Council
of Science and Technology: Washingont, DC, 1963; pp 43-45.
(18) McKelvey, V. E.; Duncan, D. C. Presented at 3rd Symp. on
Development of Petroleum Resources of Asia and Far East (ECAFE),
Tokyo, Japan, 1966.
(19) Bartlett, A. A. Math. Geol. 2000, 32 (1), 1–17.
(20) Al-Jarri, A. S.; Startzman, R. A. J. Pet. Technol. 1997, 49 (12),
1329–1338.
(21) Ivanhoe, L. F. World Oil 1995, 216, 77–88.
(22) Foreman, N. E. World Oil 1996, 217 (2), 78–83.
(23) Mohr, S. H.; Evans, G. M. Nat. Resour. Res. 2008, 17 (1), 1–11.
(24) Ivanhoe, L. F. World Oil 1996, 217, 91–94.
(25) Wattenbarger, R. A.; Maggard, J. B.; Villegas, M.; El-Banbi, A.
Advances in the Economics of Energy and Resources; Moroney, J.R., Ed.;
J.A.I. Press: Greenwich, CT, 1997.
(26) Wattenbarger, R. A. Advances in the Economics of Energy and
Resources; Moroney, J.R., Ed.; J.A.I. Press: Greenwich, CT, 1994; Vol. 8,
pp 111-133.
(27) Wattenbarger, R. A. World Oil 1994, 215, 91–97.
(28) Campbell, C. J.; Laherrere, J. H. Sci. Am. 1998, 78–83.
(29) Deffeyes, K. S. Oil Gas J. 2002, 100 (46), 46–48.
(30) Szklo, A.; Machado, G.; Schaeffer, R. Energy Policy 2007, 35,
2360–2367.
(31) Al-Fattah, S. M.; Startzman, R. A. Presented at the SPE Eastern
Regional Meeting, Charleston, WV, October 1999; Paper SPE 57463.
(32) Al-Fattah, S. M.; Startzman, R. A. J. Pet. Technol. 2000, 52 (5),
62–72.
B
Nashawi et al.
tion trend does not exhibit many major fluctuations over
time. However, recent studies31-35 have shown that most
worldwide oil producing countries display more than one
Hubbert production cycle. Therefore, the application of the
conventional Hubbert model to these countries is not ap-
propriate and does not yield good forecasting results. The
additional production cycles are apparently the result of
many factors, reflecting the state-of-the-art technological
evolution in the oil industry, government regulations, eco-
nomic conditions, and political events. The single-cycle
Hubbert model does not consider the effects of these factors.
Al-Fattah and Startzman31,32 were the first to modify
Hubbert model to account for the various factors affecting
production. They introduced what is called “multicyclic
Hubbert” approach to forecast world natural gas produc-
tion. Their model was highly accurate in generating past
production history and providing accurate forecast.
The objective of this study is to apply the multicyclic
Hubbert approach to forecast the world oil production after
evaluating production trends of 47 major oil producing
countries, which virtually supply most of the world conven-
tional crude oil. Each country will have its own prediction
model; then, the world model will be determined by combin-
ing the models of all countries. In this manner, the world
model will provide a realistic production forecast because it
covers the combined performances of all producing countries.
Estimates of the ultimate recovery, remaining reserves, and
future production of crude oil for each country will be pro-
vided. Furthermore, a separate model for OPEC countries will
be presented and the organization impact on the world oil
production and reserves will be illustrated.
2. Methodology
2.1. Mathematical Formulation. Al-Fattah and Startzman32
presented the multicyclic Hubbert equation as
( )
X k X k
e -aðt -tmax Þ
qðtÞ ¼ qðtÞi ¼ 4ðqmax Þi ð1Þ
i ¼1 i ¼1 ½1 þ e -aðt -tmax Þ 2 i
where k is the total number of production cycles, qmax and tmax
are the peak oil production rate of each cycle and its correspond-
ing time, respectively, and a is a constant. The parameters
involved in eq 1 can be calculated using a nonlinear least-squares
computation technique.
The multicyclic Hubbert model as represented by eq 1 has
several advantages over the conventional single-cycle Hubbert
equation. The most important advantages are summarized as
follows:
(1) It is derived from physical and mathematical concepts.
(2) It uses readily available historical production data.
(3) It can match, with good accuracy, historic data fluctua-
tions influenced by technological, economical, and poli-
tical events.
(4) The predictions can be easily regenerated when new
production data are available.
(5) Equation 1 has three unknown parameters; whereas the
conventional Hubbert equation has four unknown vari-
ables. This fact renders the historical production data
easier to model.
Each production cycle of the multicyclic model has its own
value of ultimate recovery, Q¥, which can be calculated from the
(33) Laherrere, J. H. Oil Gas J. 2000, 98 (16), 63–73.
(34) Mohr, S. H.; Evans, G. M. Nat. Resour. Res. 2009, 18 (1), 1–5.
(35) Al-Bisharah, M.; Al-Fattah, S.; Nashawi, I. S.; Malallah, A.
Presented at the SPE Middle East Oil and Gas Show and Conference,
Manama, Bahrain, March 2009; Paper SPE 120350.
Energy Fuels XXXX, XXX, 000–000 : DOI:10.1021/ef901240p
32
following equation:
4qmax
Q¥ ¼ ð2Þ
a u ðqact - qcal Þ2
The total ultimate recovery, Npa, is then computed by adding the
ultimate recovery from each production cycle as follows:
X k
4ðqmax Þi
Npa ¼ ð3Þ
i ¼1
ai
The logistic curve representing the cumulative production of
the multicyclic model is expressed as
X k
4ðqmax =aÞi
Q ¼ ð4Þ
i ¼1
½1 þ e -aðt -tmax Þ i
The future recoverable oil, NFR, is determined by subtract-
ing the cumulative production (eq 4) from the ultimate recovery
(eq 3).
NFR ¼ Npa - Q ð5Þ
2.2. Peak Oil Production. Three different analytical proce-
dures can be used to calculate the peak oil production. These
procedures are summarized as follows: (1) correlating back-
dated discovery data with production data with shifted time lag,
(2) using known ultimate recovery; if it is unknown, it can be
calculated by adding the cumulative production and proven
reserves, and (3) using the method of inflection points.
It is always desirable to use more than one of the above
procedures to improve the confidence level of the prediction. A
detailed derivation of the method of inflection points is pre-
sented in the Appendix.
2.3. Production Data Acquisition. Unlike other statistical
forecasting methods, the multicyclic Hubbert model uses readily
available historical production data. The data used in this
investigation were collected from many reliable sources, such
as the Twentieth Century Petroleum Statistics,36 Oil and Gas
Journal37 (OGJ) database, World Oil Journal38 (WOJ), Energy
Information Administration39 (EIA), and OPEC40 official In-
ternet database Web site. The data collected for each country
consist of annual oil production from the time data are pub-
lished up to year 2005 and recently updated reserves. Production
data of 47 major producing countries virtually holding most of
the global oil reserves were used to develop the forecasting
models.
2.4. Data Analysis Procedure. A total of 47 forecasting
models, each representing one of the investigated countries,
were developed. Historical production data for each country
were scrutinized, and the number of production cycles was
determined on the basis of initial data examination. At a later
stage of the modeling process, additional cycles where some-
times required to have a more accurate fit. The multicyclic
model as presented by eq 1 was solved using a nonlinear
least-squares numerical computation technique with initial
guesses for the unknown parameters. Each production cycle
has three parameters, a, qmax, and tmax, that need to be calcu-
lated. For example, if the historical production data of a cer-
tain country exhibit two production cycles, then six parameters
would be required for the model. The optimum values of the
parameters were obtained by minimizing the root-mean-square
of the errors, RMSE, of the production rates. RMSE is a
measure of the data dispersion around zero deviation. It is
(36) Twentieth Century Petroleum Statistics, 54th ed.; DeGolyer &
McNaughton: Dallas, TX, 1998.
(37) International Energy Statistics Sourcebook, 8th ed.; OGJ Energy
Database, PennWell Publishing Co.: Tulsa, OK, 1998.
(38) World Oil J. 1996, 32-34.
(39) Energy Information Administration, EIA, Internet Home Page
Address: http://www.eia.doe.gov/ (accessed Jan 2008).
(40) Organization of Petroleum Exporting Countries, OPEC, Inter-
net Home Page Address: http://www.opec.org/ (accessed Jan 2008).
C
Nashawi et al.
defined as
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
uP
u n
ti ¼1
RMSE ¼ ð6Þ
n
where qact and qcal represent the actual and calculated produc-
tion rates, respectively, and n is the number of data points.
Once the optimal values of the model parameters for each
production cycle are determined, the ultimate recovery for each
production cycle is computed using eq 2. The total ultimate
recovery is then calculated by summing up the ultimate recov-
eries of all production cycles as illustrated in eq 3. The cumu-
lative production is computed using eq 4 by adding the annual
crude oil production from previous years.
2.5. Model Assessment Criterion. The goodness of fit of the
different countries’ models is appraised using the coefficient of
variation factor, CV, which is defined as the ratio of the root
mean squares of errors, RMSE, of each country’s model to its
peak production rate
RMSE
CV ¼ ð7Þ
qmax
The values of CV are sorted and the percentile ranking
probabilities are then calculated using the following equation:
r - 0:5
FðCVÞ ¼ ð8Þ
n
where r is the country’s rank in the goodness of fit of the
production model and n is the total number of countries.
The goodness of fit of all models is arbitrarily classified as
excellent, very good, good, or poor. A certain country’s model
has an excellent fit if its coefficient of variation, CV, is less than
the mean, μ, of CV of all countries minus their standard
deviation, σ, (CV < (μ - σ)). This criterion indicates that the
model accurately fits the data. If the CV of the country’s model
falls between the mean minus one standard deviation and the
mean plus one standard deviation, (μ - σ) e CV < (μ þ σ), this
implies that the fit is very good. If the CV is between the mean
plus one standard deviation and the mean plus three standard
deviations, (μ þ σ) e CV < (μ þ 3σ), this indicates that the fit is
good. If the CV is greater than or equal to the mean plus three
standard deviations, CV g (μ þ 3σ), this designates that the
model is poor, and the generated values have a great deviation
from the actual data.
3. Analysis of Prediction Results
This section presents the analysis of production forecasts
for 47 most influential countries around the world in terms
of ultimate crude oil reserves. We essentially investigated
every country around the globe that has a proven oil
reserves higher than 0.468 BSTB. Furthermore, we classi-
fied the countries into two major categories: (1) OPEC
countries and (2) non-OPEC countries. First, we will pre-
sent prediction results for each category; then, we will
aggregate all the production data and provide an analysis
forecast for the entire world. Moreover, we will select Saudi
Arabia from the OPEC countries and the U.S. from the
non-OPEC countries as the most prominent countries in
each category to illustrate the multicyclic approach. Be-
cause of the large number of investigated countries and the
length limitation of the paper, it is not possible to display
the prediction model of each country in graphical form;
thus, only those countries having a prominent impact on the
supply and demand of crude oil will have their prediction
models presented in graphical form, the results of the
remaining countries will be given in table form.
Energy Fuels XXXX, XXX, 000–000 : DOI:10.1021/ef901240p Nashawi et al.

Figure 1. Saudi Arabia crude oil production model. Figure 2. Iran crude oil production model.

3.1. OPEC Countries. OPEC comprises Algeria, Angola,

Indonesia, Iran, Iraq, Kuwait, Libya, Nigeria, Qatar, Saudi
Arabia, United Arab Emirates (UAE), and Venezuela.
Eventhough Angola joined the organization in 2007; its
production will be included in the analysis since the fore-
casting period will extend beyond 2007. Before presenting
the prediction results, it is important to give a brief historical
background about the major events in OPEC history and its
influence on global crude oil production, oil price, and policy
making.
OPEC’s actual production was mainly unrestricted until
the 1973 Arab oil embargo. After that, the production
flattened at around 30 MMSTB/D until 1979, when the
Iranian revolution occurred. The average price of the U.S.
crude oil increased from $12.64/bbl in 1979 to $21.59/bbl in
1980 and then to $31.77/bbl in 1981 because of a shortage in
the crude oil supply during the Iran-Iraq war.41,42 OPEC
countries attempted to maintain high oil price by controlling
Figure 3. Indonesia crude oil production model.
the oil production when the demand fell, as a consequence of
this strategy, the oil production dropped from 30 MMSTB/ reserves and moderate energy policy. The forecasting model
D in 1979 to about 16 MMSTB/D in 1985. This move was illustrated in Figure 1 indicates that Saudi Arabia’s produc-
supposed to raise the crude oil prices; however, the reduction tion rate is anticipated to increase from 9.4 MMSTB/D in
in OPEC production was counterbalanced by an increase in 2005 to 12.2 MMSTB/D in 2015, and it is estimated to peak
production of non-OPEC countries, such as Mexico, China, in 2027, at a production rate of 14 MMSTB/D. The ultimate
U.S. (Alaskan fields), Canada, and Western European oil recovery (Npa) predicted from the model using eq 3 is
countries (North Sea fields), along with few OPEC countries about 370.5 BSTB. The oil produced by 2005, calculated by
that did not abide by their production quota. This behavior eq 4, is 107.7 BSTB. Thus, our proposed model predicts an
resulted in a sharp decrease and fluctuation in the oil price. ultimate reserve, NFR, of 262.8 BSTB (eq 5). This value
The average price of the U.S. crude oil decreased from closely matches the 262.3 BSTB reserve value published by
$24.09/bbl in 1985 to $12.51/bbl in 1986. Up to 2005, OPEC the EIA39 and OPEC.40
continues the post-1985 production trend where the produc- The prediction models of the other OPEC counties are
tion increased from 16 MMSTB/D to 31.1 MMSTB/D. As of displayed in Figures 2-12. This investigation shows that
2005, OPEC’s share of the world crude oil production was Iran (Figure 2) and Indonesia (Figure 3) have already
about 45%. reached their production peaks in 1974 and 1977, respec-
Figure 1 displays the actual and predicted oil production tively. Angola (Figure 4) and Algeria (Figure 5) will peak in
for Saudi Arabia. It is well-known that Saudi Arabia has the 2010 and 2012, respectively. The other OPEC countries will
largest oil reserve in the world, and it will remain by far the peak between 2017 and 2036. Iraq (Figure 6) is expected to be
highest crude oil producer for years to come. Consequently, the last OPEC country to reach its peak production of 8.5
there is no doubt that it will continue to be a major key player MMSTB/D in 2036 because of its huge oil reserves and the
in balancing the world crude oil market, not only because of political events that affect its production capability. More-
its high production capability but also because of its spare over, some OPEC countries other than Saudi Arabia, such as
Iraq, Iran, Kuwait (Figure 7), UAE (Figure 8), Venezuela
(41) Zaid, A. M.; Whitman, D. L. Presented at SPE Hydrocarbon (Figure 9), Nigeria (Figure 10), and Libya (Figure 11) show a
Economics and Evaluation Symposium, Dallas, TX, March 1989; Paper promising future oil production increase. Moreover, this
SPE 18916.
(42) Talabani, S. Presented at SPE Middle East Oil and Gas Show and study shows that Iran has the potential to boost its produc-
Conference, Manama, Bahrain, March 2005; Paper SPE 92888. tion to the prerevolution level of about 6 MMSTB/D.
D
Energy Fuels XXXX, XXX, 000–000 : DOI:10.1021/ef901240p
Figure 4. Angola crude oil production model.
Figure 5. Algeria crude oil production model.
Figure 6. Iraq crude oil production model.
According to our projection, this may occur around year
2032.
The oil production model of all OPEC countries is illu-
strated in Figure 13. This model was generated by combining
the multicyclic Hubbert models of all 12 OPEC countries.
Figure 13 shows an excellent match between the actual and
predicted production rates. The model indicates that OPEC
crude oil production will peak at 53 MMSTB/D in 2026. The
production is expected to decrease to 29 MMSTB/D by 2050.
At that time, the production will come from few OPEC
E
Nashawi et al.
Figure 7. Kuwait crude oil production model.
Figure 8. UAE crude oil production model.
Figure 9. Venezuela crude oil production model.
countries, mainly, Saudi Arabia, Iraq, Iran, Kuwait, UAE,
and Venezuela. Other OPEC countries will struggle to lift
output, with production dropping in Qatar (Figure 12),
Algeria, and Indonesia. This indicates that several OPEC
members such as Indonesia, Angola, Algeria, and Qatar will
soon convert from exporters to potential importers.
OPEC production model also fits the cumulative produc-
tion data very well as illustrated in Figure 14. Our proposed
approach shows that OPEC ultimate conventional crude oil
reserves and future recoverable oil, that is, oil remaining to
Energy Fuels XXXX, XXX, 000–000 : DOI:10.1021/ef901240p
Figure 10. Nigeria crude oil production model.
Figure 11. Libya crude oil production model.
Figure 12. Qatar crude oil production model.
be produced, are about 1321 BSTB and 909 BSTB, respec-
tively. Production in OPEC countries, especially in the
Middle East, is expected to increase more rapidly than in
other regions because their resources are much larger and
their production costs are generally lower. These projections
are broadly adequate with proven reserves, since OPEC oil
prices, production policies, and national policies on devel-
oping reserves are extremely uncertain. On the basis of the
2005 OPEC production of 31.1 MMSTB/D and the remain-
ing reserves, OPEC crude oil reserves are being depleted at an
annual rate of 1.25%.
F
Nashawi et al.
Figure 13. OPEC crude oil production model.
Figure 14. OPEC crude oil production versus cumulative produc-
tion.
3.2. Non-OPEC Countries. This category consists of 35
countries. The proven reserves of these countries range from
0.468 BSTB in case of Romania to 74.436 BSTB in case of
Russia. In the subsequent discussion, we will only focus on the
main producing countries that have and will have, for some-
time to come, major influence on the supply and demand of
crude oil. Consequently, we will inspect crude oil production
from U.S., Canada, Mexico, China, Brazil, Norway, India,
United Kingdom (U.K.), Russia, and Kazakhstan.
United States crude oil production reached its peak of 9.66
MMSTB/D in 1970, after that time the production started to
steadily decline. This trend was reversed by the discovery and
production from the Alaskan fields. The contribution of
Alaskan production increased from about 0.5 MMSTB/D
in 1977 to about 3 MMSTB/D in 1985. This resulted in a
short-term production increase in the U.S. Figure 15 displays
the multicyclic model as applied to the U.S. historical produc-
tion data. The model fits the daily production data remark-
ably well, as well as the cumulative production data as illus-
trated in the figure. Moreover, the multicyclic model predicts
that the total ultimate oil recovery (Npa) is 217.7 BSTB (eq 3);
188 BSTB of this volume has been produced by 2005, as
calculated by eq 4. Thus, the oil remaining to be produced, as
determined from the model (eq 5), is 29.7 BSTB. This value is
in excellent agreement with the published reserve value of 29.9
BSTB. Our calculation shows that the U.S. crude oil reserve is
being depleted at an annual rate of 6.3%.
Energy Fuels XXXX, XXX, 000–000 : DOI:10.1021/ef901240p Nashawi et al.

Figure 15. U.S. crude oil production model. Figure 17. Brazil crude oil production model.

Figure 16. Kazakhstan crude oil production model. Figure 18. India crude oil production model.

The models of Kazakhstan, Brazil, and India are displayed

in Figures 16-18, respectively. Kazakhstan (Figure 16) has a
good potential to be a key provider of crude oil to the
international market for some time to come because of its
considerable crude oil reserve of about 39.62 BSTB. As of
2005, the production rate of this country was around 1
MMSTB/D. The multicyclic model shows that Kazakhstan
has the capability to rapidly lift up its oil production to reach
its peak of 5.56 MMSTB/D in 2020.
Brazil’s model is displayed in Figure 17. Our prediction
indicates that Brazil, which has a proven reserve of about 12
BSTB, is on the verge of peak production in 2010 with a
production rate of about 2 MMSTB/D. Figure 18 shows that
the Indian historical oil production had several sharp fluc-
tuations before reaching a production plateau in 1995 at a
rate of 0.67 MMSTB/D. The multicyclic model predicts that
the Indian oil production will peak in 2015 at a production Figure 19. Russia crude oil production model.
rate of 0.94 MMSTB/D.
As of year 2009, our study shows that Russia (Figure 19), six previously listed countries only Russia is able to substantially
Mexico (Figure 20), Canada (Figure 21), Norway (Figure 22), increase its production to peak at 11.2 MMSTB/D in 2009.
U.K. (Figure 23), and China (Figure 24) have already reached The model of all 35 non-OPEC countries is displayed in
their peak production times as presented in Table 1. Both Figure 25. The figure shows an excellent match between the
Norway and U.K. production trends show steep decline in oil historical and predicted data. The model reveals that non-
production since 2000. Mexico’s production behavior sharply OPEC countries have already reached their peak production
increased from 0.42 MMSTB/D in 1970 to peak in 2004 at 3.4 rate of 39.6 MMSTB/D in 2006. The cumulative production
MMSTB/D. Since 1970, China’s production trend has steadily data predicted by the model nicely capture the actual pro-
increased from 0.2 MMSTB/D to reach its peak of 3.82 duction rates, as illustrated in Figure 26. This figure also
MMSTB/D in 2009. Our forecast demonstrates that out of the shows that the ultimate crude oil reserve of non-OPEC
G
Energy Fuels XXXX, XXX, 000–000 : DOI:10.1021/ef901240p
Figure 20. Mexico crude oil production model.
Figure 21. Canada crude oil production model.
Figure 22. Norway crude oil production model.
countries is about 819 BSTB. According to our projection,
starting from 2005 there are about 252 BSTB remaining to be
produced. This implies that non-OPEC crude oil reserves are
being depleted at a rate of 5.6% per year.
3.3. Total World Forecast. The world model comprising
all 47 countries, OPEC and non-OPEC, is displayed in
Figure 27. The model indicates an excellent match with the
actual data for both production rate and cumulative produc-
tion. As shown in the figure, the model reveals that the world
production will peak in 2014 at a production rate of 79
H
Nashawi et al.
Figure 23. U.K. crude oil production model.
Figure 24. China crude oil production model.
MMSTB/D, and then it will start declining to reach about
30 MMSTB/D in 2050. Figure 28 illustrates the world actual
production rate versus predicted cumulative production. Our
model estimates the world ultimate reserve at 2140 BSTB of
which 979 BSTB has already been produced by 2005. Thus,
the remaining oil to be produced is about 1161 BSTB. OPEC
countries hold about 78% of the world remaining reserves. On
the basis of 2005 world crude oil production and recovery
techniques, the world crude oil reserves are being depleted at
an annual rate of 2.1%. The distribution of the ultimate and
future recoverable oil of the top 11 countries in the world is
displayed in Figure 29. These countries contribute to 76% of
the total world ultimate crude oil.
Its is important to mention that the 2009 total world oil
production is about 85 MMSTB/D, our model predicts a
production rate of 74 MMSTB/D, a difference of about 11
MMSTB/D. This is basically because of three facts: First, we
did not use oil production from every oil producing country
around the world simply because the production rates of some
countries are irrelevant compared to other countries and do
not have major impact on the global crude oil supply and
demand. Second, our predictions are solely based on conven-
tional crude oil production and proven reserves, that is, crude
oil production from shale oil and oil sands is not included in the
forecast. Shale oil and oil sands production highly fluctuates
with crude oil prices, global economy, and available technol-
ogy. Third, we only used confirmed conventional crude oil re-
serves; from time-to-time, some countries provide information
Energy Fuels XXXX, XXX, 000–000 : DOI:10.1021/ef901240p Nashawi et al.

Table 1. Results of World Crude Oil Production

peak production reserve

country tmax year qmax (MMSTB/D) proven reserve (BSTB) Npa (BSTB) NFR (BSTB) RMSE (MMSTB/D) CV (%)
Algeria 2012 2.202 12.27 26.057 12.278 0.0874 3.969
Angola 2010 1.533 8 14.627 8.909 0.0521 3.399
Indonesia 1977 1.686 4.3 25.000 4.315 0.0557 3.304
Iran 1974 6.022 136.27 194.981 136.897 0.3476 5.772
Iraq 2036 8.494 115 144.581 115.053 0.3298 3.883
Kuwait 2033 6.579 101.5 138.341 101.513 0.2724 4.140
Libya 2023 4.129 41.464 66.074 41.464 0.1225 2.967
Nigeria 2017 5.286 36.22 61.085 36.219 0.1791 3.388
Qatar 2019 1.141 15.207 22.877 15.215 0.0357 3.129
Saudi Arabia 2027 13.970 262.3 370.538 262.875 0.6479 4.638
UAE 2030 6.726 97.8 122.473 97.811 0.0926 1.377
Venezuela 2028 5.487 80.012 137.324 80.141 0.1531 2.790
Argentina 1998 0.847 2.320 11.355 2.320 0.0139 1.641
Australia 2000 0.700 4.045 9.988 3.671 0.0460 6.571
Azerbaijan 2013 1.788 7 9.433 7.477 0.0222 1.249
Brazil 2010 1.981 11.772 20.947 12.368 0.0323 1.630
Brunei 1979 0.501 1.12 4.638 1.128 0.0147 2.934
Canada 1973 1.742 16.5 38.582 16.499 0.1024 5.878
China 2009 3.816 18.25 48.539 16.938 0.0422 1.106
Columbia 1999 0.838 1.453 7.913 1.511 0.0208 2.482
Congo 2000 0.265 1.784 3.514 1.781 0.0043 1.623
Denmark 2004 0.389 1.275 3.020 1.275 0.0061 1.568
Ecuador 2007 0.580 5.06 8.800 5.060 0.0218 3.759
Egypt 1993 0.893 3.72 12.821 3.720 0.0321 3.595
Equatorial Guinea 2011 0.563 1.765 2.278 1.762 0.0075 1.332
Gabon 1996 0.370 2.205 5.421 2.202 0.0138 3.730
India 2015 0.940 5.919 12.664 5.921 0.0260 2.766
Italy 2006 0.119 0.731 1.667 0.715 0.0007 0.588
Kazakhstan 2020 5.563 39.62 43.880 39.689 0.0411 0.739
Malaysia 2004 0.762 4.2 9.423 3.905 0.0267 3.504
Mexico 2004 3.383 13.67 46.990 13.747 0.1033 3.054
Norway 2001 3.418 9.691 30.600 10.366 0.0750 2.194
Oman 1998 0.900 5.572 13.469 5.593 0.0207 2.300
Peru 1985 0.190 1.097 3.310 1.023 0.0039 2.053
Romania 1976 0.299 0.468 5.866 0.467 0.0113 3.767
Russia 2009 11.153 74.436 137.722 74.304 0.2236 2.005
Sudan 2015 1.887 6.405 6.986 6.401 0.0204 1.081
Syria 1995 0.623 3 7.526 2.991 0.0103 1.653
Thailand 2007 0.115 0.527 0.845 0.518 0.0038 3.304
Trinidad & Tobago 1978 0.234 0.809 4.180 0.824 0.0100 4.274
Tunisia 1984 0.120 0.681 1.925 0.649 0.0070 5.833
Turkmenistan 2004 0.185 0.546 1.453 0.546 0.0029 1.567
United Kingdom 1999 2.649 3.998 25.663 3.540 0.0609 2.299
United States 1970 9.659 29.9 217.748 29.686 0.1384 1.433
Uzbekistan 1995 0.103 0.594 1.052 0.575 0.0034 3.301
Vietnam 2004 0.397 3.119 4.447 3.119 0.0086 2.166
Yemen 2001 0.457 2.85 5.012 2.860 0.0109 2.385
OPEC 2026 53 1321 909
non-OPEC 2006 39.6 819 252
total world 2014 79 2140 1161

on suspected new discoveries and accordingly update their
reserves based on these speculations. Unless confirmed, the
new reserves overestimate the future forecast.
3.4. Statistical Analysis. In general, our forecasting models
accurately match the actual oil production of all investigated
countries. The prediction results along with the corresponding
error analysis results of RMSE and CV values for all countries
are manifested in Table 1. Figure 30 displays a semilog plot of
CV versus the percentiles rankings. The figure shows that the
CVs are log-normally distributed around their mean, μ, of 2.85
(μ = 2.85%) and standard deviation, σ, of 1.41 (σ = 1.41%).
According to the model assessment criterion set in this study,
the models of eight countries out of the 47 investigated
countries have excellent fit with the actual published data
(CV < 1.44), thirty-three models provide very good predictions
(1.44 e CV < 4.26), six models have good predictions (4.26 e
CV < 7.08), and none of the models resulted in poor prediction
(CV g 7.08). The goodness of fit can be also illustrated by
I
inspecting the cross plot of the actual versus predicted reserves
(Figure 31), which shows that the plotted points are tightly
grouped along the unit slope line indicating excellent correla-
tion between the actual and predicted values.
3.5. Model Validation. The accuracy of the prediction
forecast is illustrated by comparing the actual production
of the last three years (2006-2008) to the predicted produc-
tion from the multicyclic models of OPEC, non-OPEC, and
total world. The production data of the last three years were
not used in the initial modeling process; they were spared for
validating the forecasting models. This is an important step
to test the reliability and effectiveness of the prediction
models. The comparison is displayed in Figures 32-34,
where the actual production rates are plotted on the same
graphs as the model predicted rates. It is quite obvious in all
three figures that the actual rates align along the solid line of
the predicted production indicating that the multicyclic
model highly duplicates the real production data. Moreover,
Energy Fuels XXXX, XXX, 000–000 : DOI:10.1021/ef901240p
Figure 25. Non-OPEC crude oil production model.
Figure 26. Non-OPEC crude oil production versus cumulative
production.
Figure 27. World crude oil production model.
we calculated the RMSE for each country for the last three
years. The models of OPEC countries have minimum, maxi-
mum, and median RMSE values of 0.068 MMSTB/D, 0.821
MMSTB/D, and 0.256 MMSTB/D, respectively; whereas, the
non-OPEC countries have a minimum RMSE value of 0.003
MMSTB/D, a maximum value of 0.924 MMSTB/D, and a
median value of 0.063 MMSTB/D. The total world model has
a median value of 0.084 MMSTB/D, which means that 50%
of the total world models deviate by less than 84 MSTB/D
J
Nashawi et al.
Figure 28. World crude oil production versus cumulative produc-
tion.
Figure 29. Ultimate and future recoverable crude oil for the world
top 11 countries.
Figure 30. Cumulative probability plot of the coefficient of varia-
tion.
from the perfect fit. The remaining 50% of the world models
have a maximum deviation of 924 MSTB/D from the perfect
match, which is less than 1.3% of the total world production.
The results of the validation analysis of all investigated
countries are presented in Table 2. A thorough inspection of
the results of each country reveals that the forecasting models
are in very good agreement with the actual production.
Energy Fuels XXXX, XXX, 000–000 : DOI:10.1021/ef901240p
Figure 31. Comparison between published and predicted world
reserves.
Figure 32. OPEC model validation.
Figure 33. Non-OPEC model validation.
4. Discussions and Conclusions
Forecasting is not accomplished by consulting a crystal ball
or a mystic of some sort, but by appraising the past, inspecting
present conditions, and projecting these into the future based
on the best available information. It is well-known that the
ultimate oil recovery of any field in the world is only deter-
mined when the production management decides to abandon
the field for good. This does not occur until the projected oil
revenues fall below expected costs and human ingenuity is
K
Nashawi et al.
Figure 34. World model validation.
unable to reverse this relationship. Therefore, it is not a sin to
acknowledge that the economic production life of any field is
almost impossible to predict efficiently. Hence, forecasts
should be flexible to adjustment whenever additional infor-
mation becomes available or as conditions change. Even
though it is inevitable to preclude the possibility of minor
inaccurate forecasting results, still it is of paramount impor-
tance that the forecast be conducted. Without the forecast, a
valid decision or public policy debate on a national or global
scale cannot be made.
Several pertinent conclusions can be drawn based on the
findings of this investigation:
(1) The multicyclic Hubbert model, which is a modified
form of the conventional Hubbert model, has been
presented and applied successfully to the prediction of
world crude oil production. The multicyclic model has
fewer parameters than the original Hubbert model
making production data easier and more accurate to
fit. Our investigation demonstrated that out of the 47
evaluated countries, 17% have excellent models, 70%
have very good models, and 13% have good models.
(2) The OPEC model was developed by combining the
individual models of the organization 12 countries.
The results of this study estimated OPEC ultimate
reserves by 1321 BSTB and the remaining oil by 909
BSTB. OPEC holds about 78% of the world crude oil.
(3) Our study also indicates that OPEC crude oil produc-
tion will peak in 2026 at a production rate of 53
MMSTB/D. On the basis of 2005 crude oil production
rate and recovery techniques, OPEC crude oil reserve
is being depleted at an annual rate of 1.25%.
(4) Despite the current world economical crisis, we specu-
late that OPEC will remain the main world supplier of
crude oil up to the end of this century.
(5) Non-OPEC countries have already reached their peak
production of 39.6 MMSTB/D in 2006. According to
our analysis, the ultimate reserve of these countries is
819 BSTB and their future recoverable oil is 252
BSTB. Non-OPEC countries hold 22% of the world
crude oil reserves, which are being depleted at an
annual rate of 5.6%.
(6) On the basis of the analysis of all 47 investigated oil
producing countries, the results of our study estimated
that the world ultimate reserve of crude oil is around
2140 BSTB and that 1161 BSTB are remaining to be
produced as of 2005 year end.
Energy Fuels XXXX, XXX, 000–000 : DOI:10.1021/ef901240p Nashawi et al.

Table 2. Results of Validation Analysis

country RMSE (MMSTB/D) country RMSE (MMSTB/D) country RMSE (MMSTB/D) country RMSE (MMSTB/D)
Algeria 0.2967 Argentina 0.0587 Equatorial Guinea 0.0569 Sudan 0.0490
Angola 0.3029 Australia 0.2994 Gabon 0.0082 Syria 0.0868
Indonesia 0.0901 Azerbaijan 0.0840 India 0.0235 Thailand 0.2327
Iran 0.1222 Brazil 0.1141 Italy 0.0066 Trinidad & Tobago 0.0117
Iraq 0.4836 Brunei 0.0138 Kazakhstan 0.1878 Tunisia 0.0252
Kuwait 0.0987 Canada 0.1266 Malaysia 0.1533 Turkmenistan 0.0784
Libya 0.1349 China 0.0231 Mexico 0.0773 United Kingdom 0.1718
Nigeria 0.8211 Columbia 0.1963 Norway 0.2886 United States 0.2318
Qatar 0.0684 Congo 0.0400 Oman 0.0554 Uzbekistan 0.0706
Saudi Arabia 0.8002 Denmark 0.0626 Peru 0.0148 Vietnam 0.0434
UAE 0.2903 Ecuador 0.0633 Romania 0.0026 Yemen 0.0143
Venezuela 0.2207 Egypt 0.0216 Russia 0.9237

RMSE (MMSTB/D)

minimum maximum median

OPEC 0.0684 0.8211 0.2555
non-OPEC 0.0026 0.9237 0.0626
total world 0.0026 0.9237 0.0840

(7) The world conventional crude oil production will peak The maximum cumulative production time is obtained by
in 2014 at a rate of 79 MMSTB/D. The world crude oil differentiating eq A-1 and setting it to zero
reserves are being depleted at an annual rate of 2.1%.
Q¥
Qmax ¼ ðA-8Þ
2
Appendix
This corresponds to the inflection point on the cumulative
production curve. The peak production rate is then calcu-
Predicting the Peak Production Using the Method of Inflec- lated by substituting eq A-8 into eq A-1
tion Points. Hubbert7 expressed the production rate q(t) or
a
dQ/dt by a parabolic equation in the cumulative production, qmax ¼ Q¥ ðA-9Þ
Q, domain as follows: 4
The peak production time tmax can be calculated by
qðtÞ ¼ dQ=dt ¼ aQ þ bQ2 ðA-1Þ transforming the logistic equation, eq A-5, into a linear form
Since the production rate will be zero when the cumulative and substituting for Qmax at tmax from eq A-8
production is equal to the ultimate recovery, Q¥, then eq A-1  
1
can be written as follows: tmax ¼ to þ lnðDo Þ ðA-10Þ
a
aQ¥ þ bQ¥ 2 ¼ 0 ðA-2Þ The two inflection points on the bell-shape curve, Hubbert
Solving eq A-2 for the constant b yields rate/time curve, are determined by taking the second deri-
vative of eq A-7 as follows:31
a
b ¼ - ðA-3Þ
Q¥ Q¥ a3 Do eaðt -to Þ ½Do 2 -4Do eaðt -to Þ þ e2aðt -to Þ 
q00 ðtÞ ¼
Substituting eq A-3 into eq A-1 and rearranging terms yields ½Do þ eaðt -to Þ 4
dQ ðA-11Þ
adt ¼ ðA-4Þ
Q2 Setting eq A-11 to zero and solving for time, t, yields the
Q- time of inflection points where changes of sign occur on the
Q¥
curve with one complete cycle. The solution gives two inflection
To express the cumulative production as a function of
points tinf1 and tinf2 defined, respectively, as follows
time, the left-hand-side of eq A-4 is integrated from to to t f
t¥ and the right-hand-side is integrated from Qo to Q f Q¥.   " #
1 pffiffiffi
This yields tinf1 ¼ to þ ln ð2 - 3ÞDo ðA-12Þ
a
Q¥
Q ¼ ðA-5Þ
1 þ Do e -aðt -to Þ and
  " #
where Do is defined as 1 pffiffiffi
tinf2 ¼ to þ ln ð2 þ 3ÞDo ðA-13Þ
Q¥ a
Do ¼ -1 ðA-6Þ
Qo
Substituting either tinf1 or tinf2 into eq A-7 yields the produc-
Equation A-5 is called the logistic equation. tion rate at the inflection points, qinf, as follows:
Differentiating eq A-5 with respect to time yields the
a
production rate equation as follows: qinf ¼ Q¥ ðA-14Þ
6
-aðt -to Þ
Q¥ aDo e
qðtÞ ¼ ðA-7Þ Using eqs A-9 and A-14, we obtain a relationship between
½1 þ Do e -aðt -to Þ 2 the maximum production rate, qmax, and the production rate
L
Energy Fuels XXXX, XXX, 000–000 : DOI:10.1021/ef901240p
at the inflection point, qinf, as follows:
3 D
qmax ¼ qinf
2 qmax = peak or maximum oil production rate of each cycle,
Nomenclature
a = constant, eq 1, 1/t, 1/year
BSTB = billion stock tank barrels
CV = coefficient of variation, eq 7, dimensionless
Do = dimensionless cumulative factor
F(CV) = cumulative probability function, eq 8
k = number of production cycles, eq 1
MSTB/D = thousand stock tank barrels per day
MMSTB/D = million stock tank barrels per day
n = number of observations, eq 6, or countries, eq 8
Npa = total ultimate oil recovery, eq 3, MMSTB
NFR = future recoverable oil, eq 5, MMSTB
q(t) = production rate, MMSTB/D
qact = actual production rate, MMSTB/D
Nashawi et al.
qcal = calculated production rate, MMSTB/D
qinf = oil production rate at the inflection point, MMSTB/
ðA-15Þ
MMSTB/D
Q¥ = ultimate oil recovery, eq 2, MMSTB
r = country’s rank in the goodness of fit of the production
trend, eq 8
RMSE = root mean squares of the errors, eq 6,
MMSTB/D
t = time, calendar year
tinf = inflection point time, calendar year
tmax = peak production time, calendar year
to = arbitrary time, calendar year
TSTB = trillion stock tank barrels
σ = standard deviation of the dimensionless root mean
squares of the errors
μ = mean of the dimensionless root mean squares of the
errors